Deep Learning Strategies for Trading Factor Model Residuals

Summary

This study replicates a deep-learning statistical-arbitrage approach that trades residuals from asset-pricing factor models. It applies the original method to a more recent out-of-sample period, using U.S. equity data from 2016 to 2024. The workflow follows the earlier study’s data preprocessing and factor modeling and uses convolutional neural networks and Transformers. The authors state that they applied point-in-time practices to avoid information leakage.

Some tests produce unusually high out-of-sample Sharpe ratios, at times above 10. The authors caution that these results may reflect overfitting, unusually favorable market conditions, or inadequate treatment of transaction costs and market impact. They call for further robustness checks and note that the results are stronger than the more modest improvements in the original research. The supplied description does not report detailed trading rules, costs, or robustness findings, limiting conclusions about real-world profitability.

Key ideas

• The strategy seeks to trade unexplained cross-sectional variation in factor-model residuals.
• The replication applies convolutional neural networks and Transformers to a more recent U.S. equity period.
• The study reports applying point-in-time principles to prevent information leakage.
• Some reported out-of-sample Sharpe ratios exceed 10, but the authors flag possible overfitting and market-specific effects.
• Transaction costs, market impact, and further robustness checks could materially change the results.

Tags

# A Deep Learning Approach for Trading Factor Residuals


# A Deep Learning Approach for Trading Factor Residuals









The residuals in factor models prevalent in asset pricing presents opportunities to exploit the mis-pricing from unexplained cross-sectional variation for arbitrage. We performed a replication of the methodology of Guijarro-Ordonez et al. (2019) (G-P-Z) on Deep Learning Statistical Arbitrage (DLSA), originally applied to U.S. equity data from 1998 to 2016, using a more recent out-of-sample period from 2016 to 2024. Adhering strictly to point-in-time (PIT) principles and ensuring no information leakage, we follow the same data pre-processing, factor modeling, and deep learning architectures (CNNs and Transformers) as outlined by G-P-Z. Our replication yields unusually strong performance metrics in certain tests, with out-of-sample Sharpe ratios occasionally exceeding 10. While such results are intriguing, they may indicate model overfitting, highly specific market conditions, or insufficient accounting for transaction costs and market impact. Further examination and robustness checks are needed to align these findings with the more modest improvements reported in the original study. (This work was conducted as the final project for IEOR 4576: Data-Driven Methods in Finance at Columbia University.)

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.